Nuclear fusion has the potential of providing an unlimited, sustainable supply of clean energy, but only if we can master the intricate physics taking place inside the reactor can we realise this great ambition.
Scientists
have been making small steps toward this aim for decades, but there are still
many obstacles to overcome. One of the biggest challenges is controlling the
reactor's unstable and super-heated plasma, but a new approach shows how we can
achieve it.
Scientists
used a deep reinforcement learning (RL) system to study the nuances of plasma
behaviour and control inside a fusion tokamak – a donut-shaped device that uses
a series of magnetic coils placed around the reactor to control and manipulate
the plasma inside it – in a joint effort by EPFL's Swiss Plasma Center (SPC)
and artificial intelligence (AI) research company DeepMind.
The coils
must make millions of small voltage adjustments per second to successfully keep
the plasma confined within magnetic fields, so it's not an easy balancing act.
 |
3D model of the TCV vacuum vessel. (DeepMind/SPC/EPFL) |
Researchers
in a new study, however, indicate that a single AI system can handle the task
entirely on its own.
"We
created controllers that can both maintain the plasma constant and be used to
correctly mould it into diverse shapes using a learning architecture that
blends deep RL with a simulated environment," the team adds in a DeepMind
blog post.
The
researchers achieved this accomplishment by training their AI system in a
tokamak simulator, where the machine learning system learned how to negotiate
the complexity of magnetic confinement of plasma through trial and error.
Following
its training period, the AI advanced to the next stage, putting what it had
learnt in the simulator into practise in the actual world.
The RL
system shaped plasma into a variety of configurations inside the reactor by
regulating the SPC's variable configuration tokamak (TCV), including one that
had never been seen previously in the TCV: stabilising 'droplets' where two
plasmas co-existed simultaneously inside the device.
In
addition to traditional shapes, the AI could sculpt the plasma into complex
configurations such as 'negative triangularity' and'snowflake' configurations.
If we can
keep nuclear fusion reactions going, each of these expressions has distinct
kinds of promise for harvesting energy in the future. The 'ITER-like shape' (as
seen above) is one of the configurations controlled by the system here, and it
may hold particular promise for future research by the International
Thermonuclear Experimental Reactor (ITER) – the world's largest nuclear fusion
experiment, which is currently being built in France.
The
magnetic control of these plasma formations, according to the researchers, is
"one of the most demanding real-world systems to which reinforcement
learning has been applied," and might pave the way for a drastic shift in
how real-world tokamaks are created.
Indeed,
some believe that what we're seeing now will have a significant impact on the
development of improved plasma control systems in fusion reactors.
"This
AI is, in my opinion, the only way forward," said Queen's University
Belfast physicist Gianluca Sarri, who wasn't involved in the work.
"There are so many variables that even a minor modification in one of them might have a significant impact on the ultimate result. If you try to do it manually, it will take a long time."
References:
- Degrave, J., Felici, F., Buchli, J. et al. Magnetic control of tokamak plasmas through deep reinforcement learning. Nature 602, 414–419 (2022). https://doi.org/10.1038/s41586-021-04301-9
0 Comments